Using fftw from MKL

[PhilippeG]

Hello

I have trouble with the fftw implemented in MKL. I have written a small code (attached)... Using gnu libfftw3 gives the right answer (mainly the output array is zero but for one coefficient).

However when linking with mkl, I do get a different (and probably false) result where all the array is filled.

Any hint would be welcomed.

[VidyalathaB_Intel]

Hi,

Thanks for reaching out to us.

>>I have trouble with the fftw implemented in MKL.when linking with mkl, I do get a different (and probably false) result where all the array is filled.

From the code which you have attached we observed that the kind value is FFTW_R2HC. 

As per the documentation, there are some known limitations of FFTW3 wrappers and their usage in Intel oneAPI MKL. The only supported r2r kind values are even/odd DFTs (sine/cosine transforms).

Please refer the following link for details regarding usage of fft wrappers

https://software.intel.com/content/www/us/en/develop/documentation/onemkl-developer-reference-c/top/appendix-c-fftw-interface-to-intel-math-kernel-library/fftw3-interface-to-intel-oneapi-math-kernel-library/using-fftw3-wrappers.html

So we suggest you to try with the supported kind value while implementing with mkl and see if it works.

You can also refer fftw examples which are installed in oneAPI directory under mkl folder.

Regards,

Vidya.

[PhilippeG · Accepted solution]

Hello

thanks for the answer. It seems that, indeed, the problem lies with the use of FFTW_R2HC. However that is really the kind of transform I need, so I guess I have to use the gnu fftw version and not MKL.

Regards

Philippe

[Gennady_F_Intel]

Philippe,

You may also try to submit this Feature Request to the Intel Online Service Center against the MKL FFT team. They collect such FR for future consideration.

[VidyalathaB_Intel]

Hi,

Glad to know that your issue is resolved.

As the issue is resolved we will no longer monitor this thread. Please post a new question if you need any additional assistance from intel.

Have a good day!

Regards,

Vidya.